This article was taken from the January issue of Wired magazine. Be the first to read Wired's articles in print before they're posted online, and get your hands on loads of additional content by subscribing online
How a Harvard maths graduate and his associates are beating Vegas and Atlantic City -- using data generated by advanced computer programs
James Grosjean -- diminutive and unshaven, dressed in jeans and a short-sleeved black shirt -- stands beneath a gaudy chandelier in the Mardi Gras-themed Showboat casino in Atlantic City, New Jersey. He focuses on a poker-based game, one of several in the place that have been designed to give the house a substantial advantage. As he watches, Grosjean explains that, even if players employ perfect basic strategy, the casino has an advantage of five per cent. He knows this because he has the data: Grosjean has devised what he says is the optimal way to play this poker variation using a piece of software that has played the game hundreds of thousands of times. "But none of the civilians play it perfectly," he says. "Most are at a 17 per cent disadvantage." Most players spend three minutes learning the ropes -- and promptly hand over their money to the casino. Grosjean has devoted hundreds of hours to studying how to give himself an edge in this game.
He turns his gaze to a shorthaired dealer from the Philippines and observes how she deals for a couple of minutes. Grosjean prefers hand-dealt games to those in which cards are distributed by machines. "Right there," he says, nodding towards one of the chairs, "is the lucky seat."
But luck doesn't come into his strategy. Grosjean, in his early forties, has a degree in applied mathematics from Harvard. He codes in half a dozen computer languages, describes himself as a statistician and says he has a mid-six-figure dollar income. At this moment, though, you would not imagine him to be educated or affluent: he wears a faded baseball cap that's positioned so the peak shades his eyes, he is missing a front tooth, looks like he hasn't seen the sun in a long while, and has avoided showering for almost a week. He resembles a punter who has gone to seed.
Playing to type, he saunters to the table, extracts a handful of bills from his pocket and buys into the game for $200. The dealer passes him 40 red chips, each worth $5. Grosjean starts betting $5 or $10 per hand. Two players materialise and snag seats to the left of him. One of them, an Asian woman, Mealea (not her real name), wagers around $300 per hand and fiddles with a Gucci purse. A burly guy wearing a football jersey has squeezed into the seat between her and Grosjean. Known as Bullet, he, too, bets small stacks of black $100 chips. Two seats to Grosjean's right sits Previn Mankodi: dark haired, bearded and with a masters in economics from Jesus College, Cambridge. He wears an Oakland Raiders T-shirt, bets small and watches intently, monitoring wins and losses.
Cards are dealt, and Grosjean starts talking in gibberish, reeling off half-thoughts about cars, fish and girlfriends. With each utterance, he is identifying the dealer's cards for Mealea and Bullet so they can place bets. The action comes lightning fast; so fast that few people in the casino -- including the dealer's boss -- realise she is flashing cards.
An experienced casino shark can gain an advantage by briefly seeing a card -- a technique known as "hole carding" -- but Grosjean's skill is to take this and apply maths to it. He's researched every statistical eventuality of the game and memorised the optimal play for each situation. Armed with data, he turns this modest act into a weapon. He reverse engineers casino games, using data to create strategies to beat them, and can memorise more than 1,000 plays at a time. When he trolls a casino, he may look like a gambler, but he's really a quantitative analyst in search of soft targets.
Back at the table, predetermined words correlate with each card's rank. "Gotta find a girlfriend," Grosjean says. "Fishin' for money here." He's just announced that the dealer has given herself a queen and a five. In case his players don't know what to do with that information, he uses hand signals. Looking relaxed and acting chatty, Grosjean spots cards, calculates optimal plays, adjusts those plays so as not to look suspicious, keeps an eye out for casino security, and signals to his teammates verbally and manually.
Leading his group through a game in which most players have no chance of breaking even, Grosjean operates with a 25 per cent advantage. With a basic strategy memorised, and with some form of card knowledge -- sometimes the exact card, sometimes a range of possibilities (ie knowing that a card is between a two and a nine) -- Grosjean calculates the perfect play. "I know more about these games than the casinos," Grosjean says later. "It's fun to figure out games and beat them. My major was applied math, and that is what I use."
Over two hours, Mealea and Bullet win $2,000 between them. The dealer remains oblivious to the plan that has just been executed, as does the pit boss -- the dealer's overseer -- and the Showboat's surveillance crew, who keep watch via overhead cameras. They ignore Grosjean in favour of more obvious targets, such as blackjack card counters who use an easily detected system to calculate the value of cards remaining to be dealt, based on knowledge of those already dealt. Typically, counters realise only a two per cent return on investment.
Grosjean works with a revolving group of players, maybe a dozen in all. He and his partner Mankodi are currently overseeing three teammates. During the coming days in Atlantic City, they'll use maths, technology and physiology to win money at games that exist solely to be unwinnable, while watching for casino managers, who routinely jettison players who discover advantages.
In April 2000, Caesars Palace security personnel in Las Vegas thought Grosjean and a collaborator were cheating by bending cards, a way of marking them. Police arrested the pair. Grosjean's mugshot is now a staple of casino-security databases, even though charges were dropped. "James Grosjean is the most intelligent advantage player out there, and his maths knowledge makes him a threat," says Willy Allison, owner of the World Game Protection Conference, an annual gathering of casino-surveillance experts. "He does nothing illegal, but he plays with an advantage. And we don't run games to lose money."
At around two the following morning, Grosjean opens the door to his motel room, a couple of blocks from the Boardwalk. The place is a dump: guests in neighbouring rooms are arguing, fluorescent lights flicker in the hallway, and the carpet feels disconcertingly sticky. Despite this, Grosjean has chosen to stash $50,000 here.
On the unmade bed there are two laptops: a Toshiba Satellite and a Sony Vaio. Data streams across both computer screens, a flow of digits enabling Grosjean to strategise plays and divine odds. The Vaio is working on numbers to help him calculate the values of "loss rebates": casinos encourage high-stakes players by offering deals to refund a certain proportion of losses. Grosjean is parsing these numbers to find an accurate value -- the games may not be beatable any other way.
Grosjean prefers to room away from the gambling tables. "Casinos are vile places, a total scam, a total con," he says. "We convey the image of being stone-cold degenerate gamblers. Except we have skills. People think I can't give this [way of life] up because I love whacking games. I do -- but I also love writing code and running numbers."
On the Toshiba, he is running "500 million hand simulations of a new game I'd rather not mention," Grosjean says. "I'm guessing our edge will degrade by a few per cent if we can't distinguish picture cards, which is often the case when I get a glimpse of paint but do not know if it is a jack, queen or king. I want to confirm how much our edge deteriorates when I'm uncertain."
Among the information on Grosjean's computers are snatched surveillance pictures of him and his teammates -- he uses the pattern on the casino carpeting to pinpoint where they have been picked off. Other files contain details of dealers who reveal hole cards (those who blatantly flash get dubbed "superstars"; those who seem flashy, but aren't, are "teases"), and phone numbers of local attorneys across the country. Grosjean has been accused -- incorrectly -- of cheating on a number of occasions. When casinos get physical, he sues; in 2005 he won a $400,000 jury verdict after he was unlawfully detained at Imperial Palace casino in Las Vegas.
The incident at Caesars Palace resulted in a jury finding the casino guilty of false imprisonment, defamation and battery;
Caesars settled out of court for an undisclosed sum.
On Grosjean's bed is a printout of a photo from the Facebook page of a player on a competing crew he suspects will be in Atlantic City over the weekend. "I like to know the species that I might encounter out here in the wild," Grosjean says with a shrug.
A life spent hacking casinos might, thanks to Hollywood, appear to be glamorous. But Grosjean views it as prosaic. "You're not seeing anything special here," he tells me. "If I was at home in Las Vegas, this is what I would be doing all day: run numbers, whack games, repeat. Monday, Tuesday, Saturday, Thanksgiving, whenever."
What Grosjean does isn't gambling; it's the implementation of statistics, combinatorics (maths dealing with combinations of objects belonging to a finite set with certain constraints), probability analysis and algebra that, on occasion, allows him to play at an advantage exceeding 100 per cent. For example, he wanted to know how much to bet on a hand of blackjack if he had discerned that the first card he'd been dealt was an ace. Standard blackjack wisdom is to bet everything -- knowing about the ace offers a 42.08 per cent advantage. Working within the Kelly Criterion (a probability theory used by investors and gamblers) Grosjean and Mankodi came up with the following formula: the Kelly-optimal bet size is a fraction f of the player's bankroll, such that f = E/V, where E and V are the expectation and variance of the gamble. "When it comes to beating casino games, James is an absolute freak of nature," says Max Rubin, a former professional gambler who's played with Grosjean and now works as a game-protection consultant for Barona Resort & Casino in San Diego, California. "But it's not just the math; it's the way he applies it. He walks into a casino and finds edges at a variety of games with a variety of methods that are unthinkable for other players. The beauty of it all is that he beats games in ways that the casinos can't figure out, and they hate him for it."
The process begins with Grosjean zeroing in on a card game in which players make decisions. "Then I write a program that plays the game," Grosjean explains. "Once I see how the game plays, I figure out how to improve my chances of winning. I have the program play with those strategies. I calculate every scenario one million times and keep working in different features until the game is profitable and simple enough to play -- you can't have 100 rules to remember and five different calculations to make on each hand. So I figure out which edges we can ignore and lose fractional percentages of advantages. Sometimes we find moves that are casino-specific. I run the numbers and see what the game will pay under those various conditions."
Grosjean glances at the digital clock on his bedside table. "Writing code for a game is hard," he says. "There are always a lot of permutations, more rules than a human can remember, and there are always things you can do to improve your edge -- which is why I keep running numbers."
It's nearing 4am, which is when dealers change shifts.
Grosjean's iPhone vibrates: team members are scouting targets around town. He heads into the night, past a petty crime scene illuminated by flashing lights from police cars. Oblivious, he walks towards the glittering casino signage.
The following afternoon, Grosjean sits on a bench near the entrance of Harrah's Resort in Atlantic City. Valet-parking attendants zip past while gamblers hustle in and out via the big glass doors. Grosjean is waiting for a play that will take place in the next 90 minutes. Killing time, he recounts wild tales: deploying a strategy that made him $1,000 an hour against a brand-new game in Indiana called Top Rung which was subsequently discontinued; emptying the chip racks on a blackjack game in Atlantic City; playing an entire shoe (eight 52-card decks shuffled together) of baccarat just so he could have a massive advantage on one big bet at the end. He recalls taking a blackjack game at MGM Grand in Las Vegas for $225,000 over the course of a single eight-hour dealer's shift. "We pounded this Chinese dealer and didn't tip her much," Grosjean says. "Then we shut down the game before another team could overplay it, get her fired and draw attention to us. We had someone tell the dealer that she was exposing cards and she got herself transferred to a different game." Grosjean grew up in Chatham Borough, a medium-sized commuter town in New Jersey. His father edited a now defunct computer magazine called Sync and worked with an author named Ken Uston. In the early 80s -- when Grosjean was around 12 -- Uston was known for writing Mastering Pac-Man, a guide to beating the arcade game. He was also the mastermind behind one of the most notorious blackjack teams in the country. "[Uston] gave me a set of autographed books on winning at video games," Grosjean says. "But I knew from my father that he was a card counter, and there was a lot of mystery around that."
By then Grosjean had his first home computer -- a Sinclair ZX80 with 1k of RAM. At high school, he learned to program in C, Pascal and Fortran. He already understood how to leverage the analytical power of computers. "Using Pascal, I created customised strategies for beating friends at mah jong," he says. "The strategies were tailored for their playing styles."
He played his first blackjack games soon after turning 21. "I was on spring break," he remembers, "and drove to Atlantic City with some high school friends. On the ride down, I taught myself basic strategy by reading a book called Winning Casino Blackjack for the NonCounter. We didn't have a big result -- plus or minus $50. But I thought it could be profitable, so I bought books on card counting and began to play."
After graduating from Harvard inthelate80sasoneofthetop60 students in his year, he enrolled at the University of Chicago and pursued a PhD in econometrics with the intention of getting a job on Wall Street. Between classes, Grosjean picked up pocket money by card counting blackjack games on the Empress riverboat casino in nearby Joliet, Illinois.
Even then he recognised that he could use data to beat the system. For example, he devised a formula that let him triple his profitability while staking just twice the money. By playing $10- rather than $5-minimum tables, Grosjean would have more games to play and thus a better chance of finding those where dealers placed the cut card deeper into the six deck shoe. Some dealers do this due to poor visible perception, others through sheer laziness -- the deeper the cut card is, the less they have to shuffle. The penetration of the cut affects the percentage of cards dealt before the reshuffle and deeper penetration offers an advantage: card counting becomes more accurate towards the bottom of a deck or shoe. Less shuffling also means more hands played, and for those playing with a small edge 100 hands per hour is better than 90.
Grosjean also looked for faster dealers and fewer players at tables, which could double game speed, increasing the amounts winnable per session. And, with more tables available, he could leave disadvantageous situations to find advantageous ones -- by observing shoes and only sitting down when the count was positive, a manoeuvre known as "wonging". Finally, at the higher-priced tables, he received less scrutiny (or "heat"): people betting $10 and higher routinely vary their bets in dramatic ways. This made him less conspicuous when playing aggressively.
But on the night of October 31, 1997, Grosjean spotted something that would make blackjack far more profitable: a dealer on the Showboat riverboat casino, in East Chicago, Indiana, had a habit of distributing cards in such a way that he could always see her down card. He nicknamed her Frontloader. Nevertheless, although Grosjean was able to harvest valuable information, the system was rarely perfect. For example, catching a glimpse of the curve of a number might indicate only that the card was a two, a six or nine. So Grosjean turned to data to further increase his advantage. "I wrote a program in C and ran simulations to find the optimal plays under uncertain conditions," he says. "I knew I could run numbers for situations in which I was unsure of what the dealer had."
Grosjean committed to memory 500 or so computer-derived strategies -- and made sure he went up only against Frontloader. "I won 57 out of 60 sessions and paid off a lot of debt," Grosjean says. "I would go to the boats every day, play heavy hours, and have my computers continually running simulations at home."
Grosjean didn't discover hole carding -- Uston published a chart for it in his book Million Dollar Blackjack in 1981 -- but he was the first player to create a strategy that accounted for frequent occurrences of partial information. Before Grosjean came on the scene, players guessed when they couldn't differentiate cards.
Grosjean was guessing too, but his calculations meant his guesses were correct a lot more often.
Around this time, Grosjean began to wonder about that career in finance. He had been in the running for a quantitative-analyst position at Goldman Sachs. A dip in the market meant the position was on hold, causing him to focus on casinos. He didn't look back.
Other games, such as three-card poker, were analysed and added to his repertoire. He made exploratory trips to Las Vegas and began putting together teams.
Grosjean published a book, Beyond Counting -- a manual of calculations and charts of new methods for beating sucker games -- in 2000 (a new edition, Exhibit CAA, sells online for $250). Card counting, which casinos knew how to spot, was analogue compared to his digital approach.
In early 2002, he teamed up with Keith Taft, a former engineer with defence contractor Raytheon. Taft, now dead, built electronic devices that assisted in beating casino games. His innovations included card-counting calculators that fitted in shoes, and computers that timed the spins of roulette wheels, to see if they biased certain numbers. He recruited Grosjean to write code for a wireless system that would allow a team of players to crush certain blackjack games. This was at a time when it was still legal in some jurisdictions to bring computers to casinos.
They looked for games with standardised shuffles that would allow software accurately to predict where certain cards would end up as long as it knew where they were at the start of the shuffle. "A player would enter the ranks and suits of cards as they were dealt, using a set of binary switches inside his sleeve," Grosjean says. "The information would be relayed to a computer, via a broadcast band, and the computer would know where the cards should be on the next shuffle. It would then provide a play that maximised expectations -- whether it was to give the [principal team member] a 21 or to bust the dealer -- based on the coming cards."
The system was two years in the making, and Grosjean recalls numerous downsides: "There were battery issues, transmission issues, issues of circuits overheating," he says. "We'd have four guys at the table playing. The inputter would feel more comfortable with someone standing behind him, to block him, so we had that person involved. Then there was another guy to whom the rig would be [discreetly] handed off in case the inputter had to get out in a hurry. The software was really complicated. I spent 18 months writing code for one stinking game! I learned a lot about blackjack but I can't say it was worth it. Casinos benefited from me spending my time on the computer rather than being out there playing."
The team played its last session on New Year's Eve 2003, hours before a law went into effect that made the use of computers in casinos illegal. "By seven that night, we had won 20-something-thousand," Grosjean says. "But we were having mechanical problems with the switches. We could have fixed them, squeezed in a last couple of hours, and risked getting caught." (Even though it was still legal, casino personnel would have frowned upon use of the computer.) "Or we could take our win, walk, and they wouldn't have a clue about what we had done. We decided to walk."
As he finishes the story, word comes that it's almost time for the play the team is making at Harrah's to be executed. Grosjean walks over to the glass doors and into the casino with the other punters.
Earlier in the day, on a scouting trip to Harrah's, team members have identified a "superstar" dealer named François (not his real name). By observing the staff-rotation patterns, Grosjean's group calculates when and at which table he'll be dealing blackjack next.
An associate, who won't be part of the play, assumes the lucky seat more than two hours beforehand -- just after 4pm -- and sits through two dealers who cover their hole cards perfectly. Mealea and Bullet monitor the table, making sure it doesn't fill up before François arrives.
As François's shift is about to begin, Grosjean swaps seats with the associate. Mealea and Bullet slide into their chairs later. To avoid possible attention, Mealea and Bullet bet $100 a hand.
Grosjean -- playing the consummate low roller to deflect exposure -- bets a measly $10.
When François arrives at the table, Bullet and Mealea triple their bets to maximise their earnings. Grosjean spots information every round and communicates the plays to his colleagues.
The scheme seems to be going well. Then Grosjean notices a brunette talking on a house phone inside the blackjack pit. She's staring down Mealea -- who's managed to win a few thousand dollars in ten minutes of moderate betting. "I need a cold soda," Grosjean says, conveying in code that the heat is on Mealea. She promptly stands up, dusts a healthy pile of chips into her purse, and disappears into the crowd. Bullet continues to play without attracting attention. The session goes well, with Grosjean spotting most of the cards, announcing their ranks for Bullet, and using hand signals to direct every play.
Minutes before the dealer's shift is scheduled to end, it is cut short. Another dealer steps in and proceeds with more caution, using his hand to block the cards from Grosjean's sight. Grosjean and Bullet continue to play with the new dealer in place, in an effort to make what they're doing less obvious, but the signs are clear -- they've been spotted by the casino. The brunette who earlier suspected Mealea sweeps over to the table. She stares at Grosjean and asks, "Is your name James?" He doesn't answer. So she asks again, more forcefully "Is your name James?"
Grosjean acts as if she's talking about the ID cards that casinos issue to players who want free rooms and dinners in exchange for gambling. "I don't have a player's card," he tells her. She walks off. Grosjean plays for another ten minutes before leaving.
That night, at a casino called Trump Marina, a copy of his mugshot from the Caesars incident is placed behind the gaming tables. In the next 24 hours, more recent pictures are distributed in Atlantic City and beyond. Grosjean is hot. Only the tightest dealers will be allowed to deal to him - if he is not thrown out altogether.
This doesn't mean the team stops its operation. He plays for another 36 hours, exploiting casinos that are unaware of his presence. Before he leaves town, he takes a midnight stroll on the Boardwalk. "There are so many casinos in the US," he says eating a slice of pizza. "On any given night, there are way more games out there than I could possibly whack."
A few days later, an acquaintance texts Grosjean, asking if he's back in Vegas, running numbers and seeking out games on his home turf. He responds that, yes, he's reverse engineering games and analysing stats. But he's circumspect about his whereabouts: "Consider me a moving target," he writes, before returning to his computers.
This article was originally published by WIRED UK
